MRI with Compressed Sensing and Variable View Sharing

A new sampling pattern enables retrospective selection of frame rates in MRI to allow for different spatiotemporal resoluiton tradeoffs and protocol optimization for different imaging applications.  Complementary poisson disk (CPD) sampling maintains near-optimal sampling for variable time windows in dynamic imaging, where the optimal spatiotemporal resolution tradeoffs are not fully known, as they depend on signal characteristics. To address this uncertainty, our proposed method uses a novel data sampling and variable data-sharing scheme with a constrained image reconstruction model to provide a flexible retrospective selection of spatial and temporal resolution. This provides robustness to motion and rapid contrast enhancement in breast and abdominal dynamic contrast-enhanced MRI. We also propose the use of a variable flip angle acquisition, which mimics these signals but offers a ground truth, to optimize spatiotemporal resolution sampling tradeoffs for specific applications.

GitHub Repository

Sampling Demo

Levine E, Daniel B, Vasanawala S, Hargreaves B, Saranathan M. 3D Cartesian MRI with compressed sensing and variable view sharing using complementary poisson-disc sampling. Magn Reson Med. 2017 May;77(5):1774-85.

Online Journal Article

Comparison of pseudorandom (DISCO), and CPD composite sampling patterns after segmenting a square region of k-space. Comlementary poisson disk (CPD) provides uniform and random sampling patterns with radially symmetric blue noise power spectra for arbitrary choices of sampling parameters.

Professor of Radiology (Body Imaging) and, by courtesy, of Bioengineering
William R. Brody Professor of Pediatric Radiology and Child Health
Professor of Radiology (Radiological Sciences Laboratory) and, by courtesy, of Electrical Engineering and of Bioengineering

Evan Levine and Manoj Saranathan are alumni of the BMR group